Standard DRAM utilizes a capacitor to store a charge. Advantages include fast read and true random access, but the device is volatile and requires refresh to maintain the stored charge. U.S. Pat. Nos. 5,995,409 and 6,222,216 describe DRAM with some contiguous memory space dedicated to nonvolatile storage. This is accomplished by “shorting” capacitors to either “1” or “0” to provide nonvolatile—but not reprogrammable—digital memory.
Flash memory provides one type of reprogrammable non-volatile memory. Flash memory read times are slow due to limitations of the floating-gate transistor—or split-channel floating-gate transistor—that forms the basis for the flash memory cell. In standard flash, in order to properly engineer the floating-gate transistor for write and erase, the transistor regions are heavily doped. This creates a high threshold voltage (approximately 1V) relative to the power supply voltage in current art (approximately 1.8V), across the drain and source. This relatively high threshold voltage is needed to avoid “unintended disturbances” in unselected nonvolatile memory cells. When 1.5V (approximately) is applied to the control gate to select the transistor, the amount of current from source to drain is accordingly relatively low, and it therefore takes a relatively long time for the circuit to drive the bit line connected to the floating-gate transistor. Thus, the time required to sense the current—which corresponds to the stored charge—is longer than sensing times in standard DRAM, for example. This challenge is further complicated and worsened for device and circuit optimization as nonvolatile memory systems demand multilevel storage (instead of binary only).
Split-channel flash (developed in the 1980s) utilizes a split-channel floating-gate transistor to provide reprogrammable non-volatile memory, but these have numerous limitations such as requiring source-side programming and the slower read times of Flash memory.